|
|
Foundations of Robotics
Seminar, April 29, 2009
Time
and Place | Seminar Abstract
Using Classification to Gain Environmental Awareness from Low-fidelity Sensors
Matt Tesch
Carnegie Mellon University - Robotics Institute
NSH 1507
Talk 4:30 pm
Classification algorithms have a wide variety of uses in image
processing, robotics applications, and many diverse fields. In this
talk, I demonstrate the use of several simple classifiers in
determining high-level environment-related information using
non-visual sensor data from an array of low-fidelity sensors. In
particular, we are interested in building hyper-redundant mechanisms
called snake robots that are able to react to their environment.
Therefore, detecting properties of their environment, such as
placement of walls or texture of the ground, is an important
requirement for the mechanisms.
The talk is composed of three distinct sections. The first will be an
basic overview of several common classification methods. Next, I will
describe the architecture of our robots, and the type of sensors we
are using. This section will conclude with a description of the
simple features we extract from these sensor measurements. The final
section will be a presentation of our results when using these
methods.
The Robotics Institute is part of the School of Computer Science, Carnegie Mellon University.